Building a PSU Dummy Load/ Test Bed

[mpilchfamily]

September 19, 2007

As Xtreme CPU grows we will be performing more and more reviews and publishing many more informative articles. In order to conduct many of these reviews there is some specialized equipment involved. This is especially true when you are reviewing a PC Power Supply (PSU). PSU testing equipment can be very costly. To have a quality testing lab you would need to spend between $5,000 and $20,000 for all the equipment. The most common equipment used for a PSU review is:

• Couple of voltage meters
• Couple of Amperage meter2
• Couple of Temperature probes
• AC Voltage conditioner 3PN1510B Variable Transformer, 120VAC Single Phase Input; 0-140VAC Output, 15A, Cased model with cord, plug, receptacle, lighted switch and fuse. - ST3PN1510B
• Oscilloscope USB Instruments
• Dummy load
• Power monitor Extech 380801 Appliance Tester AC DC Power Analyzer
• Sound monitoring

The single most expensive item is the Dummy load. These alone can cost between $2,000 and $15,000 depending on the one you choose. There are many adjustable loads you can buy that are general use and stay at the lower end of that spectrum. Then there are those made specifically for testing PSUs and cost $10,000+. So just getting start in PSU reviews take a bit of capitol.

It’s because of this large startup cost that I’ve been spending the past couple of weeks researching alternatives to find a way to get started without dropping thousands of dollars. I have discovered that I can easily build an adjustable dummy load for less then $500. At first I couldn’t believe that this was possible considering the prices of commercial equipment. But it makes since. One of the things that make the commercial equipment so expensive is there digital displays with built in meters. They are also computer controlled adjustable loads.

While conducting research into the methods used by other major site for testing PSUs I found that a couple of sites use home built loads. Sites like EOC and SPCR use home built load. Though SPCR has a lower end commercial unit they have added home made loads to increase the capabilities of there test bed. So I started taking note of the parts they where using.

SPCR Power Supply Test Rig, v.4 (and v.4.1) | silentpcreview.com
EXTREME Overclocking - Power Supply Testing Methodology Article - Page: 3 - Tweaking PC Hardware To The Max

Before I could go too far into this I had to take a step back. It’s been about 10+ years since I learned basic electronics so I had to go and re teach myself basic Parallel circuits. Once I had that all figured out I went about calculating how each site split up there loads and what parts they used. SPCR mentions what resistors they used to create one of there expansion loads and EOC offers up good pictures of there system. Both of these helped out quite a bit in verifying my math and selecting what parts I would use for the upcoming build. I won’t go into detail about the math as that would be a whole article on it own. I also used the articles from the other major review sights to help me figure out what range of load I wanted to be able to apply to PSU to offer a good testing range.

So after several days of research I decided I wanted the load tester to focus mainly on the +12v rails. It will offer the most flexibility in the +12v rail load adjustments and offer only a low medium and high load options for the 3.3V and 5V rails. Keeping the 3.3V and 5V rails set for 3 different loads helps to keep the overall cost of the test bed down. I tried to have these preset loads reflect common loads most systems would put on those rails. The specs of the Dummy Load are as follows:

• 3.3V= ~6A, ~10A, ~16A Loads
• 5V= ~5A, ~10A, ~15A Loads
• 12V1= ~2A to ~30A adjustable in 2A increments
• 12V2= ~2A to ~30A adjustable in 2A increments
• 12V3= ~2A to ~30A adjustable in 2A increments
• 5VSB= ~2A
• -12V= ~0.5A

As you can see the 5VSB and -12V rails will only use a single small load. This is pretty much a standard of the ATX12V specs. Total capacity of the dummy load will be over 1200W allowing me to test any PSU on the market today. There will be enough flexibility in the arraignment of the connectors to account for any PSU regardless of the number of actual or current limited rails it has.

Now that I had the specs I wanted to use I could go about calculating the wire wound resistors I would need to use to create those loads. Another factor I had to consider was the amount of wattage each resistor would be taking and make sure I choose a resistor that was able to cope with the heat that would be involved with that power. This was the hardest part of the whole planning process. I also wanted to keep the tolerance of the resistors as low as possible. The goal is to use resistors that have a tolerance of +/-1% but there are a few that are +/-5%. This is more then expectable and I don’t expect the loads to bee too far off from there calculated values.

In reality I was figuring out the resistors needed at the same time I was figuring out the final specs. There where more ideal conditions and loads I would have liked to use but finding the resistors to meat those specs and to be reasonably priced forced me to make many changes to the specs. There was no way I was going to pay upwards of $30 for a single resistor and have to use 8 of them. I would have liked to make the 12V loads be adjustable in 1A increments but the price would have shot up. It would have also been nice to have more flexibility in the 3.3V and 5V loads as well.

But here is the breakdown of the resistors that will be used in each bank and arraignment of the connectors.

12V Banks:

• Resistors:
• 4x HS75 2R0 F 2 Ohms 75W
• 1x HS50 3R0 1% 4 Ohms 50W
• 1x RH0256R000FC02 6 Ohms 25W
  
• Switches:
• 6x 108-0040-EVX

See the schematic pictured below.

5V Bank:

• Resistors:
• 1x HS25 1R0 1% 1 Ohm 25W
• 1x HS50 0R5 5% 0.5 Ohm 50W
  
• Switches:
• 2x 108-0040-EVX

See the schematic pictured below.

3.3V Bank:

• Resistors:
• 1x HS25 0R56 5%0.56 Ohms 25W
• 1x HS50 0R33 5% 0.330Ohms 50W
  
• Switches:
• 2x 108-0040-EVX

See the schematic pictured below.

-12V and 5VSB:

• (5VSB) 1x RH0502R500FC02 2.5 Ohms 50W
• (-12V) 1x HS10 20R 1% 20 Ohms 10W

See the schematic pictured below.

As you can see by the picture things are not as complicated as it would seam. For this reason I’ve been loosing sleep over weather or not I’m getting all my math right. But no matter how many times I go over it, it all looks good.

As you can see 12V1 and 12V2 both have an 8 pin P4 connector. This allows me to separate the P4 from the 24 pin Main if a dual rail PSU places those on separate rails. I’ve also got enough PCI-e power connectors to accommodate any configuration needed. The 2Ohm gives me 6A, 3Ohm 4A and 6Ohm 2A. So you can see buy selecting different combinations of resistors by way of the switches how I’m able to achieve between 2A and 30A loads on a single bank with only 6 resistors. This also allows you to see how the 3.3V and 5V rails work as well.

Locating these resistors wasn’t all that hard. At least not compared to trying to find the different connectors I need. I searched and search and couldn’t find what I needed. I came across the Molex web site and searched through it but couldn’t find the exact parts I needed. Then I remembered an article I’ve read and have links too in the PSU guide that explains each connector and even lists the Molex part number for them. To I signed up with Molex and requested product samples of each so that I could 1, verify I had the right connectors and 2, get at least one of each free. I figure as long as I can get a single P4, 24 pin main, and PCI-e the rest would be easy. Female 4 pin Molex and Female PCI-e connectors are easily found on adapter cables and could be salvaged from there.

As for putting the whole thing together into a working dummy load I would still need a case, wires, fans, HSs, and rivets. For the cabling I will be using wire from old PSUs I have lying around the house. I think I’ll double up the wires to help reduce resistance from the wires and prevent overheating of the wires. Insulation will have to be stripped to further guard against a fire hazard.

Heat will be a major concern inside the dummy load. If I have a 700W PSU fully loaded those resistors will be giving off allot of heat. For this reason I chose wire wound resistors that are housed in there own aluminum HS. This may not be enough so I need to have them on additional cooling. I happen to have some scrap 2mm thick aluminum plating sitting around that I may used as a HS for the resistors. I’ve also considered tearing apart all the old HDDs I have around and using there aluminum bodies to fix the resistors too. I will also need a couple of fans in the casing to help get that heat out and possibly pump it into the PSU being tested to help simulate real world conditions for the PSU.

Now I come to the case all of this will be housed in. I have an old Apex DVD play. When I say old I mean this thing is the size of an old school VCR. So there is plenty of inside room to fit everything. A bonus is the DVD drive that was in the unit used a standard 4 pin Molex power connector and it was an ATA drive. I tried to replace it but it turned out the processor on the main board was the problem not the drive. But anyway the power supply in the unit has a Molex on it, so the power supply stay to power the fans and any built in meters I may add to the unit latter. This helps so that these items are not powered by the PSU being tested which could possibly through off the expected loads being placed on the unit. Once the build gets started there will be more details about the layout of everything inside the case.

For now this is just the plan. The cost of resistors alone is about $175. This is money I do not have. Mike and Brent have offered to help fund this project but I do not expect them to foot the bill on there own. I will be putting some of my own money too this project as well as the many resources I already have around the house. I will soon be excepting donations to help make this a reality. I can’t wait to get started on this and gather the other equipment I need for the test so that I may beguine PSU reviews exclusively for XCPUs.

I look forward to hearing everyone’s comments and suggestions on this project. I will keep you all up to date on the progress of it and any technical changes made to it. I feel it is important to mention that I have taken steps to ensure that my schematics and ideas here are copy write. Even if it is a poor mans copy write. I have documented dates and everything as you can see above.

[randomizer]

So this is what you've been working on, nice. You forgot one reason why the commercial dummy loads are so expensive. Few individuals can build such specialised equipment so the companies can make a nice profit selling them knowing that there isn't really a practical alternative. It always comes down to money...

[mpilchfamily]

I think its more a factor of demand then the ability of people to build such an item. The demand for such equipment is quite low and top quality parts with controllers can be pricey plus the integrated meters.

Armed with a basic knowledge of Parallel circuits and Ohms law anyone can build one of these. Below is a basic model of one of the 12v banks. First we'll break it down resistor by resistor. Using homes law stating that A=V/R (correct labels would be I=E/R) we can see that on the 6 Ohm resistor we have a 2A load, The 3 Ohm gives us 4A and each of the 2 Ohm resistors gives us 6A. With a Parallel Circuit you break it down one resistor at a time like that then add up all the amperage values to get the total amperage involved. So you can see we have a max amperage of 30A. As we close each switch next to each resistor we can very that load. If you label the switches from 1 to 6 starting at the bottom you can see home the combinations listed below work.

Switch Combos:

• S1= 2A
• S2= 4A
• S3= 6A
• S1+S3= 8A
• S2+S3= 10A
• S3+S4= 12A
• and so on

Then the only other factor is heat disipation. If i have 12v going through a 6 Ohm resistor that gives me 2A so W=AV (P=IE) gives me 24W on that resistor. So as long as that resistor is rted at 24W or better we are in good shape.

[randomizer]

Interesting indeed. Keep us updated as to your progress

[mpilchfamily]

Well I have added an aluminum plate to the front face of the case. I don't know if I'll have the connectors on that front face or on the top of the case. I do know the switches will be on the top of the case. Also the top of the case might be modified so that the PSU being tested sits over a vent on top to take in the hot air of the testbed. Might even build a bit of a box for the PSU to sit in on the top. That way if the PSU has a fron fan rather then a bottom one it will still be taking in the hot air. I may do like what jonneyGuru does and pump the heat from the Loader through a toob and into an old case.

[Clue69Less]

Looking good! I'll try to go over it with a fine tooth comb later as I just had time to skim the details right now. I built a simple load tester a long time ago and the main thing I did wrong initially is to work too close to the limits of switches, wire and resistors. After repeated failures, I bit the bullet and upgraded. So make sure that you'll be operating with some current headroom.

Edit: Do you have access to teflon-insulated wire? It's great for projects like this. Pricey!

[mpilchfamily]

I was hopeing to get some heavy gage solid copper wire from work. There has been allot of construction lately and they have been running allot of new wire. So i should be able to get some scraps that will be long enough to solder the one side of the risistors together. Like i said i'll use doubled up or even trippled up cables from old PSUs to run to the switches since i'll need some flexability there. If i can avoid buying wire i will. Otherwise i'll go to Depo or Lows and get what i need. The switches i'm useing will be rated for 10A each. With the way everything is set up there won't be more then 10 A going through any single switch. I may have to change the 5V bank so that i have 3 1 Ohms rather then the 1 Ohm and the .5 Ohm. Since the 0.5 ohm on the 5V rail will give me 10A.

[MeltDown]

Be careful with high DC currents in switches if you're using switches that are primarily AC rated (like household light switches for example) - DC is harder on the switch contacts. Overall it looks like a good simple load. I've got a lot of active loads at work but no cameras = no PSU reviews using that equipment for me. I've thought of building my own load bank out of spare parts I have in the garage but never seem to have time. Anyway, cool project and have fun.

Are you going to get a USB O-Scope like the one jonnyguru uses? The one he's got is a good value but there are others that perform better. That one will probably work fine for measuring PSU ripple.

[mpilchfamily]

As much as i would love a full blown O-scope, one of the PC versions will have to do. SPCR also uses the same C O-scope don't know about the other major sites. My biggest problem is finding a place to put all of this stuff. Like Jonny my reviews will be done from the kitchen table.

[Blunc]

if possible I really like to have 25% headroom on high current loads and switches.

I have Digikey and Mouser accounts, if you PM an address to me I can order parts and have them shipped directly to you. I don't have tons of money but I like this project and will support it as much as I can.

Have you considered using relays with high current capabilities rather that switches that would have to carry high current? I realize this may complicate the circuit a little but it would probably add to the robustness of the tool.

Picture files don't show up here (at work) so I can't see if you've used fuses or breakers (with LED indicators) in the circuits. Knowing the status of the tool is invaluable.

[mpilchfamily]

No fuses or breaker involved. If you can view the pictures at a web page then check out the link to EOC i have up top.

Here is a basic discription of the cuircute. The Negative side of the Voltage runes down one side that is connected to one end of the resistors. The other side of the resistor is connected to a switch. Each resistor in the chain has a switch on it and then the possitive lead is connected through the other pole of all the switches. This way as a switch is turned on current is passed through the resistor thus creating a load. Since i know the value of each resistor which means i know the amperage that is allpied there i know what the total amperage is based on what switches i have turned on. I think a SPST 10A switch on each resistor will do the trick with no problem. The only LED used will be there to indicate the system (PSU being tested) is on.

Complicating the cuircuit will drasticly increase the total cost. As i said the risistors allone will cost a total of $175 + Shipping. The the switches run about $3 each. If you check the links next to the parts list you'll see i was looking at parts coming from Mouser.com. As mentioned above i may integrate a couple of amp meters into the system to show actuall amperage. But an amp meter will be used in testing to check amperage weather i have one inegrated or nor. Chances are if i integrate one it wouln't be as accurate as the seperate handheld model will be.

BTW the OP has been updated with a few links to the other peices of equipment required for PSU reviews. As i find more i'll add the links so we all have a better idea of total cost for cunducting PSU reviews.

[mpilchfamily]

**+****12V*****-
**|************|
**|************|
**|--WWW----\---|
**|**2Ω********|
**|************|
**|--WWW----\---|
**|**2Ω********|
**|************|
**|--WWW----\---|
**|**2Ω********|
**|************|
**|--WWW----\---|
**|**2Ω********|
**|************|
**|--WWW----\---|
**|**2Ω********|
**|************|
**|--WWW----\---|
**|**4Ω********|
**|************|
**|--WWW----\---|
*****6Ω********
That should give you a good idea without seeing the pictures.

--\-- Switch
--WWW-- Risistor
*** filler to make it look right.

[Blunc]

can the 2ohm resistors be 50w?

Digikey has 2 ohm 1% wirewound for a little over $5 each.

I'm going to shop the parts list at Digikey to see if I can improve on the cost, can you edit the OP to include minimum values needed for each part?

[mpilchfamily]

The 2 Ohm have to be able to take 72W at least. The 2 Ohm resistor on a 12v line will receive 72W so a 75W is needed. So unfortunatly they will have to cost about $11.50 per resistor.

The wattage rating listed is the minimum value of each. Well some can be a few watts smaller but they don't make resistors rated at odd values like 72W.

You can easily do the math to figure out the values. First take the voltage and divide it by the resistor. That gives you the amperage that it would offer. Now multiply that amperage by the voltage and you have the wattage rating needed.

[mpilchfamily]

One avenue i have not exsplored is using multiple smaller and lower wattage resistors together to equal one of the ones i need. If it would be cheaper to get say 2x 1 Ohm resistors rated at about 40 or 50W to equal one of the 2 Ohm resistors then lets go that rought. It really doesn't matter as long as i can combine them to get the volues i'm looking for. We are going for lowest cost while still maintaining the specs mentioned above to build a quallity DC Load.

[Blunc]

wouldn't you need better than 5a SPST switches for the 2-ohm loads?

Digikey has 50w al-housing resistors for $5.21 ranging from less that 1-ohm up to 5k-ohm.

I just found this: RH050, 1%, RHD 1.0 ~ 1.0K RHRD-(Value)-ND 4.63(each), 4.17(10 or more)

[mpilchfamily]

I'm going with 10A rated SPST switches.

[Clue69Less]

I can throw some bucks into the project if you want to make a quality upgrade, like better resistors or switches. I feel that having better consistency in the load between the different stages is worth the money. Do you feel like taking the time to compare prices? I may be able to send you some teflon-insulated wire too. As you get closer, keep me in the loop.

[mpilchfamily]

Sounds great. Its safe to say I'm still in the process of tweaking all the stats. Everyone's input on this project is highly valued. The wire would be great. Yes the next step is tracking down the best price and seeing if i can get all the Resistors with a tolerance of +/-1%. As of right now with the selection at Mouser.com there is only a couple of resistors that are +/-5% tolerance. Hopefully in the next few days to a week I'll be able to compile a complete parts list with links to the cheapest i could find while maintaining top quality and there prices. I'm kind of in limbo right now waiting to hear back from Molex about the product samples. I'm kinda hoping to have a box just show up out of no where. If that isn't the case then I'll buy the connectors directly from Molex unless i can find another supplier. The cheap rout i think would be to cannibalize adapter cables to get the connectors i need and just have them on cables hanging out of the box. What i really want to do is have a PCB made up and solder the connectors to it and have the PCB mounted on or inside the box. I do have a recourse for getting a professional PCB made. I've got the software all i have to do if get it all mapped out and send it in.

[mpilchfamily]

For those interested in helping me find the cheapest options for quality resistors here is the list of resistors needed and the desired specs.

All Resistors need to have a tolerance of +/-1%. The wattage rating listed for each resistor is the minimum acceptable. Higher wattages will bu much more preferred as it will help to alleviate any overheating problems on the resistors. But higher wattage ratings come at a higher cost.

2Ω 72W
4Ω 36W
6Ω 24W
1Ω 25W
0.5Ω 50W
0.33Ω 3.6W
0.56Ω 19.5W
2.5Ω 10W
20Ω 7.2W

Switches need to be DC SPST rated at 10A.